Principle Component Analysis

I'm performing PCA on a sample from my dataset (using princomp) containing 5 attributes and 20,000 values for each attribute. I then want to classify the remainder of the dataset having taken out the least important attributes. Using PCA is it possible, using the ouputs that princomp provides, to remove one or two of the attributes that are less important for classification purposes? Or does PCA not even provide this information?

Answer 1

[Andrew Newell]

If your data are in the matrix X and ndim is the number of attributes you want to retain, then you can use

[residuals,reconstructed] = pcares(X,ndim);

then reconstructed is what you want.